Method for taking rolling hyperspectral and panchromatic images combined by means of a spatial imager

Abstract

The invention relates to a method for taking rolling images by means of a spatial imager comprising, in its focal plane, a pixel-array detector the pixels of which are arranged in rows and columns, said array detector being equipped with a global shutter. According to the invention, hyper-spectral images of a region of interest are captured in various spectral bands, the hyper-spectral images being obtained by acquiring successive hyper-spectral frames, by means of said array detector; panchromatic images of the region of interest are captured, the panchromatic images being obtained by acquiring successive panchromatic frames, by means of said array detector, the acquired panchromatic frames having overlaps; the hyper-spectral frames and the panchromatic frames being acquired in alternation, the hyper-spectral frames are processed to sum the successive rows of the same strip in the hyper-spectral region.

Description

[0001]METHOD FOR TAKING COMBINED HYPERSPECTRAL AND PANCHROMATIC SCROLLING VIEWS USING A SPATIAL IMAGER FIELD OF THE INVENTION The present invention relates to a method for taking rolling views using a spatial imager comprising, in its focal plane, a matrix sensor of pixels arranged in rows and columns, said matrix sensor being equipped with an electronic shutter operating in global or rolling mode. PRIOR ART Depending on their technical characteristics, sensors on board satellites record the radiation reflected or emitted by objects on the ground in given wavelength ranges or intervals. Spectral resolution is the ability of the sensor to distinguish electromagnetic radiation of different frequencies. The more sensitive the sensor is to fine spectral differences, namely to narrow wavelength intervals, the higher the spectral resolution of the sensor. Spectral resolution depends on the deviceoptical filtering that decomposes the captured energy into more or less wide spectral bands. Panchromatic images are obtained from the recording of radiation in a single, wide wavelength interval, typically greater than 400nm. Since the data is only acquired in a single channel, only grayscale images can be obtained. If the image is coded on 8 bits, it will be viewable in 255 grayscale levels. Although less rich in terms of spectral resolution, the panchromatic image offers greater spatial resolution and signal-to-noise ratio. Hyperspectral images are obtained by sensors capable of recording information in a multitude of spectral bands, often more than 100, which are narrower, in practice of the order of a few nm, and often contiguous, in the visible, near-infrared and mid-infrared portions of the electromagnetic spectrum. The dataHyperspectral sensors therefore provide more detailed information on the spectral properties of a scene and allow more precise identification and discrimination of objects than broadband sensors. Each pixel of a hyperspectral image contains the information collected in large acquisition windows spread over the entire visible and infrared spectrum. The amount of information to be stored and processed is therefore significant, and requires more substantial computing capacity than in the case of multispectral images. The applications of hyperspectral imaging are numerous. Among the most important are geology, precision agriculture, forestry, aquatic environment management, or security. According to the prior art, hyperspectral bands are produced by filters deposited as close as possible to the pixels along lines. A hyperspectral band covers h_Hs lines, the panchromatic band covers h_Pan lines of thedetector or another detector. To increase the signal-to-noise ratio (SNR), the photons converted into electrons are summed from pixel to pixel along a column over several successive lines. The SNR is then increased by a root factor (N) with N the number of summations. The summation, from line to line, can be internal to the detector, if the detector is designed for it, or external in dedicated reading and processing electronics, as in the patent document published under number CN109640012A. This technique is called scrolling image acquisition by acquisition of lines summed along the columns or so-called pushbroom image acquisition with TDI (Time Delay Integration). The pixels of a detector convert the received photons into electrons. These electrons are accumulated in wells specific to the pixels. These wells have an accumulation capacity limited to a saturation threshold (“Full Well Capacity” inEnglish language) in number of electrons beyond which accumulation is no longer possible. The signal corresponding to acquisitions presenting saturations is typically considered degraded or even unusable. In the case of a sensor having panchromatic pixels and hyperspectral pixels having the same saturation threshold, for the same observed scene, the greater width of the panchromatic spectral band causes a faster filling of the wells of the panchromatic pixels compared to the wells of the hyperspectral pixels. It is therefore necessary, in order to maximize the number of charges integrated in the panchromatic pixels and in the hyperspectral pixels, while avoiding the saturation phenomenon, to adopt a specific strategy. To avoid saturation of the panchromatic pixels, one can either attenuate the signal of the panchromatic channel by using for example a density, or have smaller panchromatic pixels. The TDI summationline already defined for hyperspectral frames is sometimes also used on panchromatic frames to increase the SNR of panchromatic images. Another technique in use consists of acquiring successive frames on a detector equipped with N spectral filters distributed in N zones following the direction of movement of the sensor (patent document published under number US2012 / 300064A1). The principle is the acquisition for each spectral band of a succession of "Snap shots" with overlap. The final strip is a succession of (2D) imagettes with overlap for each spectral band. This type of shooting imposes a condition on the height of each spectral filter in order to be able to join the successive imagettes. The number of spectral bands is limited by the height of the sensor and the height of the filters [Number of bands = H detector / H filter]. This acquisition technique is adaptedto multispectral instruments, number of spectral bands limited to 10. It is not suitable for hyperspectral instruments (more than 50 bands), concerned by the invention. SUMMARY OF THE INVENTION Considering the prior art set out above, a technical problem that the invention proposes to solve is to carry out, in the same scrolling image capture, the acquisition of hyperspectral frames with line summation (TDI), and panchromatic frames so as to obtain hyperspectral and panchromatic images of the same area of ​​interest, while maximizing the quantity of charges integrated by the pixels by independent adjustment of the panchromatic and hyperspectral exposure time. The panchromatic frames acquired in successive “snap-shots” are used to enrich the hyperspectral image. The proposed solution of the invention to this technical problem has as its first object a method of taking scrolling images by means of a spatial imager comprising, in its focal plane, atat least one matrix sensor of pixels arranged in rows and columns, said matrix sensor being provided with an electronic shutter, characterized in that it comprises the following steps according to which: hyperspectral shots are taken of an area of ​​interest according to at least two spectral bands, the hyperspectral shots being obtained by the acquisition of successive hyperspectral frames, by means of said matrix sensor associated with a summation processing on at most h_HS successive lines of the same band in the hyperspectral area; panchromatic shots are taken of the area of ​​interest, the panchromatic shots being obtained by the acquisition of successive panchromatic frames, by means of said matrix sensor, the acquired panchromatic frames having overlaps; the acquisition of the hyperspectral frames and the acquisition of the panchromatic frames being carried out alternately; the alternating acquisitionsare carried out, while maintaining a fixed reading rate, by replacing a hyperspectral frame with a panchromatic frame; the number of successive hyperspectral frames acquired between two panchromatic frames being greater than or equal to the number of summed hyperspectral lines. Advantageously, - a summation processing of successive panchromatic frames is carried out in common areas resulting in imagettes which have a minimum overlap height between them, the imagettes being joined to form a final panchromatic product; - a gain and offset correction is carried out on the hyperspectral summation in alternating acquisition mode, and in that, for this correction, the procedure is as follows: let I be the level generated by an ideal pixel, which has no offset and a unit gain, during observation of the scene to be measured, or nHS the number of frame intervals describing the acquisition patternalternating hyperspectral and panchromatic, let kp ^ [[1,nHS]] be the index of the interval in which a PAN frame was substituted for an HS frame, let S = [[l, nHS]]\kp be the set of HS frame indices involved in the TDI summation, let (0 k ,g k ) the offset and gain of the pixel of index k € [[1,nHS]], these quantities being measured by calibration acquisitions in the laboratory and in flight, a level measured by the pixel of index k € S is Zk = I.gk + Ok the result of the sum is: Z the application of the correction of the sum is carried out in such a way as to find an ideal level I: ; - a selection is made on the panchromatic frames, the selection of said frames being read on the detector(s) and, once read, saving said selection in a memory; - the selection reduces the volume of data to be processed or transmitted, while maintaining performance in terms of coverage of the ground area to be acquired; - the temporal frequencies at which vibratory phenomena presented by the line of sight of the instrument are sampled are chosen; - the hyperspectral acquisition is combined with a super-resolved panchromatic acquisition; - instead of summing the panchromatic frames to improve a signal-to-noise ratio, said frames are used to improve the resolution by carrying out measurements of the inter-frame shifts and recombining them to improve the resolution; - the hyperspectral image is enriched by acquisitions on at least one panchromatic area of ​​height h_pan lines at a ratesufficient, greater than 10 frames per second, and a video of the scene observed is acquired during the travel time of the h_pan panchromatic lines; - the sensor comprises at least two panchromatic zones and at least one hyperspectral zone; - exposure times of the panchromatic zones are set so as to obtain a high dynamic range acquisition; - information contained in the acquisition of a first panchromatic zone to be traveled from the zone of interest is used to adjust the exposure time and / or the zone of interest and / or the application of a binning of the corresponding frames of the following panchromatic zone in the order of scrolling on the zone of interest to be observed, to optimize the image quality, and / or adjust the exposure time and / or the zone of interest and / or the application of a binning of the frames corresponding to the hyperspectral zones, to optimize the image quality; - video information is enriched by having two successive videosof the same objects on the ground separated by a time gap between panchromatic acquisitions of the two panchromatic zones, a video obtained by a first panchromatic zone visualizing moving objects in an observed scene, a video of a second panchromatic zone providing information on a followed direction; - the hyperspectral acquisition is combined with a panchromatic acquisition in stereovision, a use of the two panchromatic zones on the detector equipped with the hyperspectral filter enriching the hyperspectral image with a stereo image of the same observed zone, a stereo angle being equal to an angular gap between two panchromatic channels multiplied by a slow motion factor of the satellite; - a grouping of pixels is carried out independently between the hyperspectral and panchromatic images; - the panchromatic images are used to improve the final resolution of the hyperspectral image through a fusion processinghyperspectral-panchromatic. The second subject of the invention is a spatial imager comprising, in its focal plane, at least one matrix sensor of pixels arranged in rows and columns, said matrix sensor being provided with an electronic shutter, for implementing the method of taking scrolling images defined above. BRIEF DESCRIPTION OF THE FIGURES The invention will be better understood upon reading the non-limiting description which follows, written with reference to the appended drawings, in which: Fig. 1 shows schematically an element of a matrix sensor for implementing the method of the invention, for acquiring hyperspectral images, comprising 3 spectral bands; Fig. 2 shows schematically an element of a matrix sensor for implementing the method of the invention, for acquiring panchromatic images; Fig. 3 shows schematically the implementation of the method according to the invention comprising two adjacent zones of the same sensor for taking imagespanchromatic and hyperspectral; Figs. 4A and 4B illustrate, schematically, the scrolling shots taken according to the invention, with summation of lines in the hyperspectral zone; Fig. 5 illustrates, schematically, a typical processing of the panchromatic frames according to the invention; Fig. 6 illustrates, schematically, the alternating hyperspectral and panchromatic shots according to the invention; Fig. 7 shows a sequence of 11 frames read on a detector, followed by a sequence of a panchromatic frame, according to the invention; Figs. 8A and 8B show a configurable strategy for selecting frames according to the invention; and Fig. 9 shows a diagram of a sensor comprising two panchromatic zones, for implementing variants of the invention. DETAILED DESCRIPTION OF THE INVENTION The invention relates to a method for taking scrolling shots. This method of scrolling shots is implemented usingof a space imager. Space imaging is a technique for observing a remote area of ​​interest based on taking images in the optical domain from space by equipment installed on board artificial satellites. The area of ​​interest is in particular a terrestrial area of ​​interest. For the image capture according to the invention, the space imager comprises, in its focal plane, one or more sensors or detectors. Since the image capture is scrolling, the image at the focal plane scrolls along the columns of the detector. The sensor is a matrix sensor composed of pixels. It is most often a CMOS (Complementary Metal Oxide Semiconductor) sensor. As illustrated in Figs. 1 and 2, the pixels of the sensor are arranged in rows and columns. The matrix sensor comprises a hyperspectral detection zone. The hyperspectral image capture is carried out according to different spectral bands. In the example of theFig. 1, the sensor has 3 pixel zones corresponding to 3 spectral bands. The shooting is then carried out according to these 3 spectral bands SB: a first spectral band SB1, a second spectral band SB2 and a third spectral band SB3. In the above example, the matrix sensor has 6 lines of pixels per spectral band SB. We therefore have h_Hs=6, for each spectral band SB1, SB2 and SB3. A hyperspectral sensor is a sensor having spectral bands of width between 0.1 and 50nm, more particularly between 2 and 20nm, and a total number of spectral bands between 2 and 2000, more particularly between 50 and 200. The matrix sensor also has a panchromatic detection zone. This panchromatic detection zone is illustrated in Figure 2. It includes a number of lines of pixels, for example several hundred lines, h_Pan. According to the invention, the number of columns of pixels in the panchromatic zone is advantageously equal to thenumber of columns in the hyperspectral area. The acquisition of the same area of ​​interest by hyperspectral bands and panchromatic bands is sought in order to maximize the amount of information acquired and enrich the final hyperspectral products. This concerns hyperspectral-panchromatic fusion, correction of line-of-sight instabilities between images of a hyperspectral cube, simultaneous hyperspectral super-resolved panchromatic acquisition, simultaneous panchromatic and hyperspectral video acquisition, simultaneous panchromatic and hyperspectral stereo acquisition, control and modification of acquisition parameters (exposure time, binning, area of ​​interest, etc.) during shooting. The term binning is used in French. It can be translated as "containerization" or "classification of data". The invention proposes to advantageously use the same sensor for the hyperspectral channel and the panchromatic channel.Thus, the panchromatic detection zone and the hyperspectral detection zones advantageously form a single unit in the sensor, as shown in Fig. 3, or two separate zones. The matrix sensor according to the invention is advantageously equipped with a global electronic shutter. The shuttering is carried out at each frame, that is to say at each reading. The acquisition of the images forming the shot is simultaneous on all the pixels. It is carried out at a frequency such that the offset between two frames is less than or equal to the Sampling Time (Frame Time) ≤ 1 / Tech with Tech = GSD (Ground Sampling Distance) x Ground Scroll Speed. This is shown schematically in Fig. 4A. The frame acquisition frequency is equal to the inverse of the exposure time added to the dead time. The exposure time is maximized to collect the maximum number of photons in each pixel without saturating the electron storage capacity. The frequencyframe acquisition is adjusted to respect the sampling time. As illustrated in Fig. 4B, hyperspectral shots are acquired in push-broom mode. They are obtained by acquiring n_HS successive hyperspectral frames, using the matrix sensor. A summation is performed from line to line within each band. To perform a summation on at most h_HS successive lines belonging to the hyperspectral bands, n_HS successive frames of the ROI_HS area of ​​interest are acquired. In the example in Fig. 4B, there are 3 spectral bands and each spectral band covers 6 lines. The succession of rectangular shapes presenting the 3 spectral bands represents the succession of sensor readings, which are performed at regular times. There are 10 readings shown in Fig. 4B. The small circles shown correspond to the same object point on the ground seen along the same bandhyperspectral and through 6 successive readings. The reference frame shown at the top left of Fig. 4B is a spatio-temporal reference frame of the sensor, in which ALT means Along Track (scrolling direction) and ACT means Across Track (direction perpendicular to the scrolling or following the swath). As shown in Fig. 5, panchromatic shots are acquired in frames. A frame is the result of reading the area of ​​interest on the sensor. The frames are successive and of height h_Pan. They have overlaps. Such acquisitions are called push-frames. The final product is an image formed from the succession of common areas, namely the succession of thumbnails joined by butt-jointing. TDI summation is performed on the common areas. It results in thumbnails that maintain a minimum overlap height between them. This overlap makes it possiblethe joining of the thumbnails to form the final PAN product of the scene to be observed. As illustrated in Fig. 6, according to the invention, hyperspectral shots of an area of ​​interest ROI_HS are taken according to HS acquisition times, and panchromatic shots of an area of ​​interest ROI_Pan are taken according to a Pan acquisition time. The hyperspectral and panchromatic scrolling shots are taken with summation. The combined hyperspectral / panchromatic scrolling shots are multi-purpose. The method according to the invention therefore proposes a mixed scrolling shot, combining the summation of lines on fine spectral bands and advanced processing of frames on wide spectral bands. The summation of lines is carried out within each spectral band. The sensor is thus commanded to alternate the acquisition of the hyper-spectral zone ROI_HS (Region Of Interest HS), with an HS exposure time, with the acquisitionin the panchromatic area ROI_Pan (Region Of Interest Pan) with an exposure time Pan. The acquisition of the n_HS hyperspectral frames and the n_Pan panchromatic frames is carried out alternately. In addition, the acquisition of at least one of the n_Pan panchromatic frames is inserted into the acquisition of the n_HS hyperspectral frames. In other words, in the example of Fig. 6, one reading is deleted every 6 hyperspectral readings to insert, instead, a panchromatic reading. This replaces a hyperspectral frame with a panchromatic frame while maintaining a fixed reading rate and maximizing the exposure time of the HS frames. It should be noted that this operation is only possible on detectors capable of performing an “Integration While Read” (IWR) type acquisition, i.e. detectors capable of parallelizing the exposure phase of frame n with the reading phase of frame n-1.number of successive hyperspectral frames acquired n_HS between two panchromatic frames is greater than or equal to the number of summed hyperspectral lines (typically equal). We therefore have one panchromatic frame for every 5 hyperspectral frames. The common areas shown in this figure are the common areas between two successive panchromatic frames. The PAN panchromatic frames to be kept are calculated to ensure overlap between the summed common Pan panchromatic areas. The overabundant PAN panchromatic frames are removed. Selection of panchromatic frames The insertion of the PAN panchromatic frame can be carried out every n_HS frames or multiple frames. The maximum hyperspectral summation number is in this case n_HS- 1. n_HS being typically much lower than the height h- Pan of the acquired PAN frames, it is possible to acquire the PAN frames only sparsely while guaranteeing complete coverage of the scene to be observed. There remains ainterest in acquiring several redundant frames. This allows TDI sums to be performed between frames in order to improve the signal-to-noise ratio of the final product. An example of a strategy for selecting panchromatic frames to be acquired is given in the following paragraphs. Fig. 7 describes a sequence of 11 HS frames read on a detector, followed by a panchromatic frame. The panchromatic frame can be saved or discarded, depending on the panchromatic frame selection strategy. seq_HSPAN(12)_savedPAN denotes a sequence for which the panchromatic frame is saved, seq_HSPAN(12)_discardPAN denotes a sequence for which the panchromatic frame is discarded. A parameterizable selection strategy is that defined by the sequence salvo_comb(n, m, p), which is illustrated in Fig. 8A. salvo(n, p) is defined by the sequence of n seq_HSPAN(12)_savedPAN between which are interspersed p seq_HSPAN(12)_discardPAN. In Fig. 8A, n=3 and p=2. As isillustrated in Fig. 8B, the sequence salvo_comb(n, m, p) is defined as the sequence containing a salvo(n, p) followed by m seq_HSPAN(12)_discardPAN. Hyperspectral image equalization correction The insertion of PAN frames has the consequence of modifying, for each spectral band, the distribution of the n_HS-1 lines that will be summed. The impact is an error on the pixel-to-pixel correction. Each pixel of the sensor, indexed by (i,j), has a response characteristic in offset o(i,j) and gain g(i,j). o(i,j) and g(i,j) are measured, in the laboratory or in flight, by calibration acquisitions. The exploitation of the raw data at the output of the instrument requires applying a correction dependent on this calibrated pixel characteristic. For processing bandwidth reasons, the correction must be applied to the TDI HS sums, and not to the individual pixels. Due to the use of the HS+PAN alternating acquisition mode, the TDI HS sums imply (n_HS-1)pixels among n_HS. The index of the pixels involved in the TDI HS sum is known a priori and allows the selection of the calibration data to be applied to the result as part of the correction. The processing of hyperspectral frames in alternating acquisition mode with panchromatic frames is made more difficult by the use of the integration and summation method with time transfer (TDI), in particular when a correction of the pixel offsets and gains must be applied to the sum. For gain and offset correction on the HS TDI sum in HS+PAN alternating acquisition mode: let I be the level generated by an ideal pixel, which has no offset and a unit gain, when observing the scene to be measured, let nHS be the number of frame intervals describing the HS+PAN acquisition pattern, let kp ^ [[1,nHS]] be the index of the interval in which a PAN frame has been substituted for an HS frame, let S = [[1,nHS]]\kp be the set of indices ofHS frames involved in the TDI summation, i.e. (0k,gk) the offset and the gain of the pixel of index k € [[1,nHS]], these quantities being measured by calibration acquisitions in the laboratory and in flight, the level measured by the HS pixel of index k € S is Z k = Ig k + O k the result of the TDI HS sum is: [Math 1] Z the application of the correction of the TDI HS sum is carried out in such a way as to find the ideal level [Math 2] I: Hyperspectral-panchromatic fusion The mixed panchromatic-hyperspectral acquisition of the same scene on the ground gives access to a final product obtained by ground processing called hyperspectral-panchromatic fusion (“pansharpening” in English). The panchromatic image containing information typically of higher spatial resolution, it is then used as a support to be “colored” using one or more narrow spectral bands. The result is then a hyperspectral image with the spatial resolution of the panchromatic image combined with a better resolved panchromatic acquisition via super-resolution processing or binning described later, we can access a hyper-resolved hyperspectral image.Correction of motion between images of a hyperspectral cube The localization on the same sensor of the hyperspectral and panchromatic bands is sought in order to improve the combined spatial registration ("co-registration" in English) and to be able to correct the line of sight shifts between the acquired hyperspectral frames. Indeed, the panchromatic frames acquired in pushframe, have intrinsically a strong geometric rigidity because all the pixels of the imagettes are acquired at the same time and the successive imagettes can be realigned simply by a correlation processing in the overlap zone between successive imagettes. Each spectral image of the hyperspectral cube is formed of successive lines which will move over time because of the instabilities of the platform.The panchromatic frames acquired alternately with the hyperspectral frames will serve as a geometric basis for resetting the hyperspectral lines. The maximum accessible resetting frequency is the acquisition period of the panchromatic frames. The resetting processing will preferably be carried out on the ground. Simultaneous panchromatic and hyperspectral video acquisition According to another implementation variant of the invention, the HS hyperspectral acquisition is combined with a video acquisition. In such a case, instead of summing the Pan panchromatic frames to improve the signal-to-noise ratio (SNR), they are used to produce a video of the observed area. The video frequency is equal to the frequency of the panchromatic frames, typically 20 frames per second (tps). The observation duration is the scanning time of the panchromatic area, typically 1 s.The video acquisition can be carried out on one or more panchromatic zones. As shown in Fig. 9, advantageously, a detector is used which comprises at least two panchromatic zones, namely a Panchromatic Zone 1 and a Panchromatic Zone 2. This makes it possible to enrich the video information by having two successive videos of the same objects on the ground separated by the time difference between the panchromatic acquisitions of Panchromatic Zones 1 and 2, which is typically 5 s. In one example, the video obtained by Panchromatic Zone 1 makes it possible to visualize the moving objects in the observed scene, while Panchromatic Zone 2 provides information on the direction followed. According to yet another alternative embodiment of the invention, a detector is used, as previously, comprising two separate panchromatic zones, namely a Panchromatic Zone 1 and a Panchromatic Zone 2.However, according to the present variant, the HS hyperspectral acquisition is combined with a panchromatic acquisition in stereovision (stereo). The use of two panchromatic zones on the detector equipped with the hyperspectral filter makes it possible to enrich the hyperspectral image with a stereo image of the same observed zone. The stereo angle B / H is equal to the angular difference between the two panchromatic channels PAN multiplied by the satellite slow motion factor. Typically, a stereo angle B / H that can be achieved is 0.12. Simultaneous hyperspectral super-resolved panchromatic acquisition According to yet another implementation variant of the invention, the HS hyperspectral acquisition is combined with a super-resolved panchromatic acquisition.In such a case, instead of summing the panchromatic frames to improve the SNR, they are used to improve the resolution by measuring the inter-frame shifts and recombining the frames and thus taking advantage of spatial sampling to improve the resolution. This implementation variant is suitable for hyperspectral imagers, which have a low aperture number that maximizes the number of photons, and thus have a high cutoff optical modulation transfer function. The resolution gain is at least a factor of 1.5. Control and modification of acquisition parameters (exposure time, binning, area of ​​interest, etc.) during shooting Due to the scrolling of the image on the sensor of the spatial imager, the panchromatic image is acquired before the hyperspectral image. A reasonable delay between these two acquisitions is from a few milliseconds to several seconds.It is suitable for real-time processing to optimize the quality of the images acquired in the following hyperspectral and / or panchromatic areas. The information from the panchromatic pixels (signal level, distribution) can be used and the sensor acquisition parameters can be modified for the following frames, according to thresholds, in order to improve, for example, the RSB by increasing the exposure time, improve the GSD by eliminating binning or the selection of panchromatic frames. Ultimately, the invention has many advantages. In particular, the invention makes it possible to: acquire very high frequency panchromatic frames, interlaced with HS hyperspectral frames. This makes it possible, by ground processing, to perform corrections for interband geometric drifts linked to satellite instabilities on the spectrum acquisition horizon.This de-constrains the long-term stability of the platform; hyper-spectral of the same area with dynamic access to panchromatic images and adapted for the panchromatic channel since the time of panchromatic acquisitions can be modified.This improves the signal-to-noise ratio of the panchromatic image or makes possible panchromatic acquisitions with high dynamic range (“HDR” in English); to obtain, in certain variants of implementation of the invention, panchromatic video images, simultaneously with obtaining the hyperspectral image; to obtain native stereo images, simultaneously with the hyperspectral image; and to increase the panchromatic resolution by ground processing, with oversampling, with the images taken simultaneously with the hyperspectral image to access intelligent operating modes making it possible to improve image quality performance by exploiting the information received by the first panchromatic zone to access better resolved hyperspectral products thanks to ground processing of hyperspectral-panchromatic fusion (“pansharpening” in English).The invention therefore makes it possible to mix on the same sensor and almost simultaneously a pushframe type acquisition on the panchromatic zone and pushbroom (with a posteriori line summation) on the zone equipped with spectral filters, while maintaining a fixed sensor reading rate by replacing a hyperspectral frame with a panchromatic frame and being able to accept at least 50 spectral bands.

Claims

CLAIMS 1. Method for taking scrolling images using a spatial imager comprising, in its focal plane, at least one matrix sensor of pixels arranged in rows and columns, said matrix sensor being provided with an electronic shutter, characterized in that it comprises the following steps according to which: hyperspectral images are taken of an area of ​​interest according to at least 50 spectral bands, the hyperspectral images being obtained by the acquisition of successive hyperspectral frames, by means of said matrix sensor associated with a summation processing on at most h_HS successive lines of the same band in the hyperspectral area; panchromatic images are taken of the area of ​​interest, the panchromatic images being obtained by the acquisition of successive panchromatic frames, by means of said matrix sensor, the acquired panchromatic frames having overlaps;the acquisition of the hyperspectral frames and the snapshot acquisition of the panchromatic frames being carried out alternately, while maintaining a fixed reading rate, by replacing a hyperspectral frame with a panchromatic frame; the number of successive hyperspectral frames acquired between two panchromatic frames being greater than or equal to the number of summed hyperspectral lines.

2. Method according to claim 1, characterized in that a summation processing of successive panchromatic frames is carried out in common areas resulting in thumbnails which have a; minimum overlap height between them, the images being joined to form a final panchromatic product. 3.Method according to one of claims 1 or 2, characterized in that a gain and offset correction is carried out on the hyperspectral summation in alternating acquisition mode, and in that, for this correction, the procedure is as follows: let I be the level generated by an ideal pixel, which has no offset and a unit gain, during observation of the scene to be measured, let nHS be the number of frame intervals describing the alternating hyperspectral and panchromatic acquisition pattern, let kp ^ [[1,nHS]] be the index of the interval in which a PAN frame has been substituted for an HS frame, let S = [[1,nHS]]\kp be the set of HS frame indices involved in the TDI summation, let (0k,gk) be the offset and gain of the pixel with index k € [[1,nHS]], these quantities being measured by laboratory and in-flight calibration acquisitions, a level measured by the pixel with index k € S is Z. k = I.gk + Ok the result of the sum is: Z = the application of the correction of the sum is carried out in such a way as to find an ideal level I: .

4. Method according to one of the preceding claims, characterized in that a selection is made on the panchromatic frames, the selection said frames being read on one or more detectors and, once read, saving said selection in a memory.

5. Method according to claim 4, characterized in that the selection reduces the volume of data to be processed or transmitted, while maintaining performance in terms of coverage of the ground area to be acquired.

6. Method according to one of the preceding claims, characterized in that a hyperspectral line recalibration processing is carried out to correct for motion.

7. Method according to one of the preceding claims, characterized in that the hyperspectral acquisition is combined with a super-resolved panchromatic acquisition. 8.Method according to one of the preceding claims, characterized in that, instead of summing the panchromatic frames to improve a signal-to-noise ratio, said frames are used to improve the resolution by carrying out measurements of the inter-frame shifts and recombining them to improve the resolution.

9. Method according to one of the preceding claims, characterized in that the hyperspectral shot is enriched by acquisitions on at least one panchromatic zone of height h_pan lines at a sufficient rate, greater than 10 frames per second, and a video of the scene observed is acquired during the travel time of the h_pan panchromatic lines.

10. Method according to one of the preceding claims, characterized in that the sensor comprises. at least two panchromatic zones and at least one hyperspectral zone.

11. Method according to claim 10, characterized in that exposure times of the panchromatic zones are adjusted so as to obtain a high dynamic range acquisition.

12. Method according to one of claims 10 or 11, characterized in that information contained in the acquisition of a first panchromatic zone to be covered in the area of ​​interest is used to adjust the exposure time and / or the area of ​​interest and / or the application of a binning of the corresponding frames of the following panchromatic zone in the order of scrolling over the area of ​​interest to be observed, to optimize the image quality, and / or adjust the exposure time and / or the area of ​​interest and / or the application of a binning of the frames corresponding to the hyperspectral zones, to optimize the image quality. 13.Method according to one of claims 10 to 12, characterized in that video information is enriched by having two successive videos of the same objects on the ground separated by a time gap between panchromatic acquisitions of the two panchromatic zones, a video obtained by a first panchromatic zone displaying objects in motion in an observed scene, a video of a second panchromatic zone providing information on a direction followed.

14. Method according to one of claims 10 to 13, characterized in that the hyper-spectral acquisition is combined with a panchromatic acquisition in stereovision, a use of the two panchromatic zones on the detector equipped with the filter. hyperspectral enriching the hyperspectral image with a stereo image of the same observed area, a stereo angle being equal to an angular difference between two panchromatic channels multiplied by a slow motion factor of the satellite.

15. Method according to one of the preceding claims, characterized in that the panchromatic images are used to improve the final resolution of the hyperspectral image through hyperspectral-panchromatic fusion processing.

16. Spatial imager comprising, in its focal plane, at least one matrix sensor of pixels arranged in rows and columns, said matrix sensor being provided with an electronic shutter, for implementing the method of taking scrolling images according to one of the preceding claims.

Images